Skip to content

[InstCombine] Only fold extract element to trunc if vector hasOneUse #115627

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 20, 2024
Merged

[InstCombine] Only fold extract element to trunc if vector hasOneUse #115627

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,9 +205,9 @@ Instruction *InstCombinerImpl::foldBitcastExtElt(ExtractElementInst &Ext) {
if (IsBigEndian)
ExtIndexC = NumElts.getKnownMinValue() - 1 - ExtIndexC;
unsigned ShiftAmountC = ExtIndexC * DestWidth;
if (!ShiftAmountC ||
(isDesirableIntType(X->getType()->getPrimitiveSizeInBits()) &&
Ext.getVectorOperand()->hasOneUse())) {
if ((!ShiftAmountC ||
isDesirableIntType(X->getType()->getPrimitiveSizeInBits())) &&
Ext.getVectorOperand()->hasOneUse()) {
if (ShiftAmountC)
X = Builder.CreateLShr(X, ShiftAmountC, "extelt.offset");
if (DestTy->isFloatingPointTy()) {
Expand Down
18 changes: 6 additions & 12 deletions llvm/test/Transforms/InstCombine/extractelement.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -722,20 +722,14 @@ define i8 @bitcast_scalar_index_variable(i32 %x, i64 %y) {
ret i8 %r
}

; extra use is ok if we don't need a shift
; extra use is not ok, even if we don't need a shift
Copy link
Contributor

@nikic nikic Nov 13, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This case could be okay as we're just replacing extractelement with trunc here, without also inserting a bitcast. But tentatively I'd agree that if we're going to cast to a vector anyway, it makes sense to stay in vector representation.


define i8 @bitcast_scalar_index0_use(i64 %x) {
; ANYLE-LABEL: @bitcast_scalar_index0_use(
; ANYLE-NEXT: [[V:%.*]] = bitcast i64 [[X:%.*]] to <8 x i8>
; ANYLE-NEXT: call void @use(<8 x i8> [[V]])
; ANYLE-NEXT: [[R:%.*]] = trunc i64 [[X]] to i8
; ANYLE-NEXT: ret i8 [[R]]
;
; ANYBE-LABEL: @bitcast_scalar_index0_use(
; ANYBE-NEXT: [[V:%.*]] = bitcast i64 [[X:%.*]] to <8 x i8>
; ANYBE-NEXT: call void @use(<8 x i8> [[V]])
; ANYBE-NEXT: [[R:%.*]] = extractelement <8 x i8> [[V]], i64 0
; ANYBE-NEXT: ret i8 [[R]]
; ANY-LABEL: @bitcast_scalar_index0_use(
; ANY-NEXT: [[V:%.*]] = bitcast i64 [[X:%.*]] to <8 x i8>
; ANY-NEXT: call void @use(<8 x i8> [[V]])
; ANY-NEXT: [[R:%.*]] = extractelement <8 x i8> [[V]], i64 0
; ANY-NEXT: ret i8 [[R]]
;

%v = bitcast i64 %x to <8 x i8>
Expand Down
22 changes: 22 additions & 0 deletions llvm/test/Transforms/InstCombine/vector_insertelt_shuffle.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,4 +90,26 @@ define <4 x float> @bazzzzzz(<4 x float> %x, i32 %a) {
ret <4 x float> %ins1
}

; test that foldBitcastExtElt doesn't interfere with shuffle folding

define <4 x half> @bitcast_extract_insert_to_shuffle(i32 %a, i32 %b) {
; CHECK-LABEL: @bitcast_extract_insert_to_shuffle(
; CHECK-NEXT: [[AVEC:%.*]] = bitcast i32 [[A:%.*]] to <2 x half>
; CHECK-NEXT: [[BVEC:%.*]] = bitcast i32 [[B:%.*]] to <2 x half>
; CHECK-NEXT: [[INS3:%.*]] = shufflevector <2 x half> [[AVEC]], <2 x half> [[BVEC]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; CHECK-NEXT: ret <4 x half> [[INS3]]
;
%avec = bitcast i32 %a to <2 x half>
%a0 = extractelement <2 x half> %avec, i32 0
%a1 = extractelement <2 x half> %avec, i32 1
%bvec = bitcast i32 %b to <2 x half>
%b0 = extractelement <2 x half> %bvec, i32 0
%b1 = extractelement <2 x half> %bvec, i32 1
%ins0 = insertelement <4 x half> undef, half %a0, i32 0
%ins1 = insertelement <4 x half> %ins0, half %a1, i32 1
%ins2 = insertelement <4 x half> %ins1, half %b0, i32 2
%ins3 = insertelement <4 x half> %ins2, half %b1, i32 3
ret <4 x half> %ins3
}


Loading